Iterative point-wise reinforcement learning for highly accurate indoor visible light positioning.

Area of Science:

• Robotics
• Computer Vision
• Wireless Communication

Background:

• Indoor positioning systems (IPS) are crucial for various applications.
• Visible Light Positioning (VLP) offers a promising alternative to traditional IPS.
• Existing VLP methods struggle with noise and height inaccuracies.

Purpose of the Study:

• To introduce Iterative Point-Wise Reinforcement Learning (IPWRL) for highly accurate indoor VLP.
• To address the challenges of non-deterministic noise and deterministic height errors in VLP.
• To develop a VLP algorithm that does not require a training phase.

Main Methods:

• IPWRL iteratively updates height information to improve positioning accuracy.
• The algorithm was evaluated in a Received Signal Strength (RSS) based VLP system.
• Performance was compared against conventional RSS, K-Nearest Neighbors (KNN), and non-iterative PWRL algorithms.

Main Results:

• IPWRL effectively mitigates non-deterministic noise and deterministic height errors.
• The algorithm demonstrated robust positioning performance across the entire tested area without training.
• IPWRL successfully corrected significant height mismatches, yielding low positioning errors compared to other methods.

Conclusions:

• IPWRL provides a robust and accurate solution for indoor VLP.
• The iterative height updating mechanism is key to its superior performance.
• IPWRL offers a viable alternative to supervised learning methods in VLP systems.